Choice of the Pseudo-Optimal Configuration of a Cooled Gas-Turbine Blade Based on a Constrained Minimization of the Global Entropy Production Rate

Abstract

The problem of determining the optimal configuration of a cooled gas-turbine blade is approached by an entropy minimization technique proposed in previous works by the same authors. The present paper describes the application of the same line of thought to a more complex (and realistic) pseudo-optimization procedure, in which the objective function is again the global entropy generation rate, but two integral constraints are added to the original formulation: the maximum blade temperature (weak constraint) and the overall enthalpy drop of the working fluid in the blade passage (strong constraint). The discontinuous optimization procedure is presented here in an application which resembles a trial-and-error technique, but can be rigorously and formally described and implemented [12]. As a “zero configuration”, a realistic 2-D geometry is considered, and the thermo-fluiddynamic field around it is computed via a standard finite-element code. Then, the entropy generation rates in the blade/fluid system are calculated, and the value of the overall enthalpy drop of the gas as well as the value and location of the maximum blade temperature are recorded. Keeping all other parameters fixed (in particular, maintaining the same cooling air flowrate), the geometry of the blade is slightly “perturbed”, by introducing arbitrary modifications in the blade profile, the number and location of cooling holes, etc. Again, the velocity and temperature fields are computed, and inlet conditions are tuned so that the overall enthalpy drop remains approximately constant and the blade maximum temperature does not exceed a certain assigned value. An “optimal” configuration is found, which is affected by the minimal entropy generation rate, while abiding to the imposed constraints. The procedure is demonstrated on a realistic blade profile, and is shown to produce a better performing cascade, at least in this 2-D simulation. The extension to 3-D problems is — in principle — straightforward (but see Section 3 for further comments).